PeakD update: discovery experiments, better SEO, and a few quality-of-life fixes

This is one of those updates where there is a mix of very visible stuff, quiet backend/metadata work, and a handful of "why was this annoying before?" fixes.  Hopefully these updates will make your daily experience better 🤞 ### A first pass at a new Discovery Feed A new **Discovery Feed** is now available from the user menu:  I want to frame this one properly: this is **not** me saying "here is the final perfect recommendation feed." Not even close. What I wanted to showcase here is more of a working concept for what we can build using the social graph and public activity already available on Hive and HAF. The current version looks at recent votes from accounts you follow, finds posts that are being picked up by that circle, and mixes in content from both familiar and unfamiliar authors. That algorithm will almost certainly change. Maybe the weight of votes needs tuning. Maybe freshness is too strong or not strong enough. Maybe we need more signals, fewer signals, better filtering, or different behavior depending on the type of account. This is exactly the kind of thing that needs real usage before it can become good. So for now, treat it as an experiment and a starting point. The important part is not this exact formula. The important part is that PeakD can start building more interesting discovery surfaces that are personal, explainable, and still based on Hive data. I also added a small "How Discovery works" modal inside the page, because if we are going to experiment with ranking, the least we can do is be clear about what is happening.  Your existing preferences still apply after loading: hidden tags, ignored accounts, NSFW settings, and minimum author reputation are respected. ### Better SEO metadata, now with JSON-LD This is probably not the flashiest thing in the release, but it is important. Kudos to @louis88 for suggesting this a while back. PeakD now adds **JSON-LD structured data** in the generated page metadata layer. That means we are not only setting the usual title, description, Open Graph and Twitter tags, but also giving crawlers a more structured explanation of what the page actually is. For example: - post pages can expose `SocialMediaPosting` data - account pages can expose `ProfilePage` data - community/profile-like pages can expose `CollectionPage` data - generic routes still get sane `WebSite` / `WebPage` fallbacks This helps with cleaner previews, better machine-readable context, and a stronger base for search/discovery outside PeakD. It is the kind of work users may not notice directly, but it matters if we want Hive content to travel better around the web. ### Witnesses, FAQ, and Support pages got some attention A few pages got some data and visible updates: - the **Witnesses** page now shows estimated witness reward stats in the header, including block reward, total daily witness rewards, and an average top-20 daily figure. Please help me double check the math here and report any issues you notice 🙏 - the **FAQ** page got a cleaner expandable layout and better dark mode handling - the **Support Us** page was reworked into a clearer action-based layout, with progress across the different ways people can support PeakD - the Snaps link in the top bar got a small redesign so it stands out a bit more ### Faster-feeling image and app loading This is probably not easily visible if you have a good network connection and a fast computer, but a decent amount of work was focused around image rendering and startup behavior. Images now get more consistent layout attributes, which shoul […]